Age of Empires 2 has long been a staple of the real-time strategy genre, celebrated for its historical campaigns and complex competitive meta, but it is now serving as the foundation for a profound philosophical and technical critique of modern artificial intelligence. A Microsoft researcher has recently utilized the game scenario editor to construct a functioning neural network powered entirely by goats. This experiment was not merely a display of technical wizardry but a deliberate attempt to demonstrate why we must stop assuming that large language models (LLMs) possess human-like consciousness simply because they have been trained on natural language datasets. By stripping away the sophisticated interface of modern AI and replacing it with literal livestock, the project highlights the mechanical and non-sentient nature of computation.
| Lead Researcher | Adrian de Wynter |
| Platform Used | Age of Empires 2 Scenario Editor |
| Logical Components | Goats and Grass |
| System Type | 1-bit Perceptron Neural Network |
| Publication Date | May 2026 |
| Primary Goal | De-anthropomorphizing LLMs |
Building a Goat Powered Logic Gate in Age of Empires 2
The technical execution of this project is as impressive as it is absurd, utilizing the intricate trigger systems found within the Age of Empires 2 editor to simulate binary logic. The researcher constructed a functional NOT AND (NAND) gate and a 1-bit perceptron, which serves as a foundational building block for more complex neural networks. In this simulated environment, bridges represent a binary 1, while grass represents a 0. The movement and placement of goats within these structures act as the bits of information being processed. This methodology mirrors how players have historically used redstone in other sandbox titles, but applying it to a rigid strategy game provides a starker contrast between the input and the resulting computational output.
By choosing Age of Empires 2 as the medium, the researcher aimed to make the underlying processes of an LLM visible and undeniably mechanical. When a user interacts with an AI like ChatGPT, the fluidity of the language often masks the mathematical operations occurring beneath the surface. However, when those same operations are visualized as goats moving across a map, the illusion of sentience evaporates. This demonstrates that the behaviors we perceive as human-like in modern AI are actually the result of complex weight relationships and operations that are, fundamentally, no more sentient than a group of digital animals reacting to game triggers.
Challenging the Anthropomorphism of Artificial Intelligence
The core thesis of the study, titled ‘If LLMs Have Human-Like Attributes, Then So Does Age of Empires 2’, uses absurdism to attack the growing tendency of both the public and the scientific community to anthropomorphize AI. The researcher argues that because LLMs generate natural language responses that sound human, we are hardwired to project consciousness onto them. By showing that the same mathematical logic can be executed by goats within a strategy game, the experiment proves that the output is independent of the ‘soul’ of the processor. If we do not believe a game of Age of Empires 2 is sentient, then logically, we should not grant that same status to an LLM performing the same calculations.
The Peer Review Crisis and Systematic Bias
This research also sheds light on a significant issue within the computer science academic field regarding how researchers approach AI. After peer-reviewing over 300 papers in the last two years, the researcher found that more than half of them began with the faulty assumption that LLMs possess human-like traits. This bias can lead to flawed conclusions in scientific research, as experiments are often designed to confirm perceived humanity rather than to understand the actual mechanics of the model. The goat experiment serves as a wake-up call, suggesting that scientists must perform experiments that see LLMs for what they are—mathematical engines—rather than what we wish them to be.
Future Implications for Game Design and AI Integration
For the gaming community, this experiment opens up fascinating questions about the future of AI in interactive media. As developers begin to integrate LLMs into NPCs and quest design, there is a risk of creating experiences that feel uncanny or deceptive. Understanding that these systems are essentially high-level logic gates, much like the goat-powered circuits in Age of Empires 2, allows players and developers to maintain a healthy distance from the technology. It encourages a shift toward using AI as a tool for emergent gameplay rather than a replacement for human-authored narrative depth. The lesson here is clear: the medium of natural language is just a skin, and the internal logic remains a series of 0s and 1s.
The Age of Empires 2 logic experiment reveals the dangerous gap between computation and consciousness
By stripping away the veneer of natural language and replacing it with tangible game objects, this study forces us to confront the reality that LLMs are mathematical engines rather than sentient beings. The use of goats as bits highlights that the medium of expression often dictates our emotional response more than the underlying logic. For the gaming industry, this reinforces the need for more transparent AI systems that prioritize mechanical reliability over simulated personality.
For further technical details on the neural network construction, you can view the official research documentation at Arxiv.
Final Pulse Score: 9.0 / 10
Related Article: History of OpenAI Five in Dota 2
Related Article: Gabe Newell Investment in OpenAI and Dota 2